facebookresearch/fairseq

This model has order 20 but KenLM was compiled to support up to 6

Opened this issue · 0 comments

What is your question?

Hi guys,When I execute the following command using the Lao 20gram language model in MMS,the following error occurs:
RuntimeError: /root/anaconda3/envs/meta/temp_dir/kenlm/lm/model.cc:49 in void lm::ngram::detail::{anonymous}::CheckCounts(const std::vector&) threw FormatLoadException because `counts.size() > 6'.
This model has order 20 but KenLM was compiled to support up to 6. If your build system supports changing KENLM_MAX_ORDER, change it there and recompile. With cmake:
cmake -DKENLM_MAX_ORDER=10 ..
With Moses:
bjam --max-kenlm-order=10 -a
Otherwise, edit lm/max_order.hh.

Code

python /root/anaconda3/envs/meta/fairseq/examples/mms/asr/infer/mms_infer.py --model "/root/anaconda3/envs/meta/fairseq/models_new/mms1b_all.pt" --lang "lao" --audio "/root/anaconda3/envs/ASRtest/audio/test/lao.wav" --extra-infer-args "decoding.type=kenlm decoding.beam=500 decoding.beamsizetoken=50 decoding.lmweight=2.69 decoding.lmpath=/root/anaconda3/envs/meta/lmdecode/lao_char_20gram.bin decoding.lexicon= decoding.silweight=2.8 decoding.unitlm=True common_eval.post_process=none decoding.results_path='/root/anaconda3/envs/meta/fairseq/0417'"

What have you tried?

cmake -DKENLM_MAX_ORDER=10 .., and recompiled the kenLM
the issue remains unresolved. What should I do?

Additionally, on the webpage https://huggingface.co/facebook/mms-cclms, the download links for the Mandarin and Cantonese language models have turned into errors. Where else can I download them from?